NEFM DNA Methylation Correlates with Immune Infiltration and Survival

Li et al. Clin Epigenet (2021) 13:112 https://doi.org/10.1186/s13148-021-01096-4

RESEARCH Open Access NEFM DNA methylation correlates with immune infltration and survival in breast cancer Dandan Li1†, Wenhao Zhao2†, Xinyu Zhang2, Hanning Lv2, Chunhong Li2* and Lichun Sun2*

Abstract Background: This study aims to determine whether NEFM (neuroflament medium) DNA methylation correlates with immune infltration and prognosis in breast cancer (BRCA) and to explore NEFM-connected immune gene signature. Methods: NEFM transcriptional expression was analyzed in BRCA and normal breast tissues using Oncomine and Tumor Immune Estimation Resource (TIMER) databases. The relationship between NEFM DNA methylation and NEFM transcriptional expression was investigated in TCGA. Potential infuence of NEFM DNA methylation/expression on clinical outcome was evaluated using TCGA BRCA, The Human Protein Atlas and Kaplan–Meier plotter databases. Association of NEFM transcriptional expression/DNA methylation with cancer immune infltration was investigated using TIMER and TISIDB databases. Results: High expression of NEFM correlated with better overall survival (OS) and recurrence-free survival (RFS) in TCGA BRCA and Kaplan–Meier plotter, whereas NEFM DNA methylation with worse OS in TCGA BRCA. NEFM transcriptional expression negatively correlated with DNA methylation. NEFM DNA methylation signifcantly negatively correlated with infltrating levels of B, CD8+ T/CD4+ T cells, macrophages, neutrophils and dendritic cells in TIMER and TISIDB. NEFM expression positively correlated with macrophage infltration in TIMER and TISIDB. After adjusted with tumor purity, NEFM expression weekly negatively correlated with infltration level of B cells, whereas positively correlated with CD8 + T cell infltration in TIMER gene modules. NEFM expression/DNA methylation correlated with diverse immune markers in TCGA and TISIDB. Conclusions: NEFM low-expression/DNA methylation correlates with poor prognosis. NEFM expression positively correlates with macrophage infltration. NEFM DNA methylation strongly negatively correlates with immune infltration in BRCA. Our study highlights novel potential functions of NEFM expression/DNA methylation in regulation of tumor immune microenvironment. Keywords: Breast cancer, NEFM, DNA methylation, Lymphocytes, Tumor-infltrating, Prognosis

Introduction Breast cancer (BRCA) is the most common malignancy among females worldwide. Clinical outcome has been improved over the past two decades with currently avail- *Correspondence: [email protected]; [email protected] able modalities, including surgery, chemotherapy, endo- †Dandan Li and Wenhao Zhao contributed equally to this work 2 Department of Breast Medical Oncology, Harbin Medical University crine therapy, radiotherapy, and targeted therapy, but Cancer Hospital, No.150 Haping Road, Nangang District, Harbin 150081, BRCA treatment remains challenging because of high China heterogeneity [1–3]. Immunotherapy is emerging as Full list of author information is available at the end of the article new therapeutics in BRCA. Several immunotherapeutic

© The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons. org/ licen ses/ by/4. 0/ . The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons.org/ publi cdoma in/ zero/1. 0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Li et al. Clin Epigenet (2021) 13:112 Page 2 of 18

agents have been explored in various tumors, includ- BRCA [14–23]. DNA methylation-mediated silencing of ing adoptive cell therapies, vaccines, oncolytic viruses, neuroflament genes (NEFH, NEFM, NEFL) is a frequent and most notably immune check point blockade (ICB). event that contributes to the development and progres- Agents of ICB such as inhibitors of cytotoxic T-lym- sion of BRCA [8]. phocyte-associated antigen (CTLA-4), programmed cell In ovine amniotic epithelium (oAECs) isolated from death receptor1 (PD-1), and programmed cell death1 late amnia, NEFM mRNA levels were signifcantly ligand1 (PD-L1) have been widely used in solid tumors, increased, while immunomodulatory efect of inhibiting refractory cancers harboring microsatellite instability lymphocyte proliferation was lost, and global DNA meth- and classical Hodgkin lymphoma. Notably, anti-PD-L1 ylation was enhanced. Myelin oligodendrocyte glycopro- antibody atezolizumab in combination with nab-pacli- tein induced incomplete tolerance of CD4 (+) T cells taxel has been approved for the treatment of metastatic specifc for myelin and neuronal self-antigen NEFM in triple-negative breast cancer (TNBC) [1–4]. Expres- mice [24, 25]. Tese studies suggest that NEFM is related sion of PD-L1 in infltrating immune cells is required for to immune response. However, the relationship of NEFM response to atezolizumab plus nab-paclitaxel in IMpas- with TILs in tumor progression or immunotherapy sion130 trial [5]. remains unclear. Tumor-infltrating lymphocytes (TILs) comprise a In this study, association between NEFM expres- mixture of cytotoxic T cells, helper T cells, B cells, mac- sion and prognosis of BRCA was explored using TCGA rophages, natural killer cells, and dendritic cells, which (Te Cancer Genome Atlas), Te Human Protein Atlas, have been observed in many solid tumors, including Oncomine and Kaplan–Meier plotter. In addition, asso- BRCA. TILs may provide prognostic and predictive clues ciation between NEFM DNA methylation and NEFM in BRCA and other cancers. To date, robust predictive transcriptional expression was analyzed using BRCA biomarkers for immunotherapy have not been estab- samples in TCGA. Moreover, the relationship of NEFM lished in BRCA [1, 6]. TILs are more commonly observed transcriptional expression and NEFM DNA methylation at higher levels in TNBC and HER2-positive BRCA com- with tumor-infltrating immune cells was investigated pared with estrogen receptor (ER)-positive and HER2- in TCGA BRCA based on Tumor Immune Estimation negative BRCA [1, 7, 8]. TILs may be associated with Resource (TIMER) and TISIDB (tumor–immune system improved prognosis and better response rates to neoad- interactions). juvant therapy [7]. Te NEFM (neuroflament medium), located on Results 8p21.2, encodes neuroflament medium polypeptide and NEFM transcriptional expression levels in various cancers assembles along with neuroflament heavy polypeptide Diferential transcriptional expression of NEFM was pro- (NEFH) and neuroflament light polypeptide (NEFL) into fled in tumor and adjacent non-malignant/normal tis- 10-nm flamentous structures, known as neuroflaments. sues of multiple cancer types using Oncomine database. Neuroflaments comprise axon skeleton functionally to NEFM transcriptional expression was downregulated in maintain neuronal caliber and participate in intracel- most cancers, including brain and CNS, breast, colorec- lular transport to axons and dendrites. Neuroflaments tal, gastric, kidney, esophageal, ovarian, head and neck, have been implicated in biopathology of neurological cervical cancers, and lymphoma, while NEFM transcrip- diseases, including MDD (major depressive disorder) tional expression was upregulated in bladder, breast, [8–11]. NEFM belongs to dopamine receptor-interacting kidney, lung cancer, and sarcoma (Fig. 1a). To explore dif- protein (DRIP) gene family, which afects multi-aspects ferential expression of NEFM between tumor and normal of dopamine receptor activity [12]. Besides, NEFM is tissues, RNA-seq data derived from multiple malignan- associated with early response to antipsychotic medica- cies in TCGA were examined by TIMER. NEFM expres- tion [13]. Importantly, NEFM is involved in tumorigene- sion was signifcantly lower in BLCA (bladder urothelial sis/carcinogenesis [14–16]. NEFL and NEFM are located carcinoma), BRCA (breast invasive carcinoma), COAD within 8p21, and LOH of this chromosome region has (colon adenocarcinoma), HNSC (head and neck carci- been described in several cancers including BRCA [17– noma), LUAD (lung adenocarcinoma), PRAD (prostate 19]. Additionally, NEFM is potentially involved in pan- adenocarcinoma), READ (rectum adenocarcinoma), creatic cancer development and progression. Moreover, STAD (stomach adenocarcinoma), KICH (kidney chro- aberrant expression and methylation of neuroflament mophobe), KIRC (kidney renal clear cell carcinoma), and genes have been detected in ovarian cancer, esophageal UCEC (uterine corpus endometrial carcinoma), com- squamous cell cancer, renal cell cancer, glioblastoma, pared with adjacent normal tissues. By contrast, NEFM neuroendocrine tumors, prostate cancer, uterine carci- expression was comparable between tumor and normal nosarcoma, Ewing sarcoma, hepatocellular cancer and tissues in THCA (thyroid carcinoma), KIRP (kidney renal Li et al. Clin Epigenet (2021) 13:112 Page 3 of 18

Fig. 1 NEFM transcriptional expression levels in diferent types of human cancers. a Increased or decreased NEFM in diferent cancers compared with normal tissues in Oncomine database. b Human NEFM transcriptional expression levels in diferent tumor types from TCGA database as determined by TIMER. (*p < 0.05; **p < 0.01; ***p < 0.001)

papillary cell carcinoma), CHOL (cholangiocarcinoma), Kaplan–Meier plotter (Table 1, Fig. 2e–r.) Notably, a ESCA (esophageal adenocarcinoma), LIHC (liver hepa- higher level of NEFM expression correlated with favora- tocellular carcinoma), LUSC (lung squamous carcinoma), ble OS of pancreatic ductal adenocarcinoma, pheochro- THCA (thyroid carcinoma) and LIHC (liver hepatocellu- mocytoma and paraganglioma, whereas with poor OS lar carcinoma) (Fig. 1b). of kidney renal clear cell carcinoma, lung adenocarcinoma, stomach adenocarcinoma, bladder carcinoma, Prognostic potential of NEFM in cancers head–neck squamous cell carcinoma, ovarian cancer Potential impact of NEFM expression on overall sur- and sarcoma as demonstrated in Kaplan–Meier plotter vival (OS) was evaluated in Pan-cancer RNA-seq in databases.

Table 1 Impact of NEFM on overall survival (OS) in Pan-cancer RNA-seq in Kaplan–Meier plotter Cancers No. of patients MST (OS,Month) HR p NEFM high NEFM low expression expression

Kidney renal clear cell carcinoma 530 36.57 52.23 1.49 0.031 Lung adenocarcinoma 501 40.3 54.4 1.55 0.0046 Stomach adenocarcinoma 371 25.97 56.2 1.5 0.025 Bladder carcinoma 405 31.37 42.33 1.35 0.047 Head–neck squamous cell carcinoma 499 37.8 66.73 1.43 0.012 Ovarian cancer 373 43.8 45.47 1.37 0.027 Sarcoma 259 36.27 82.13 1.96 0.0011 Pancreatic ductal adenocarcinoma 177 22.03 15.57 0.57 0.011 Cervical squamous cell carcinoma 304 29.3 68.4 1.52 0.13 Lung squamous cell carcinoma 495 44.87 71.1 1.33 0.059 Rectum adenocarcinoma 502, NA NA 3.17 0.098 Thyroid carcinoma 502 NA NA 0.36 0.053 Pheochromocytoma and paraganglioma 178 NA NA 0.18 0.035 Uterine corpus endometrial carcinoma 543 103.73 51.6 0.7 0.097 Signifcant p value < 0.05 is in bold Li et al. Clin Epigenet (2021) 13:112 Page 4 of 18

(See fgure on next page.) Fig. 2 Kaplan–Meier survival curves of high versus low expression of NEFM in TCGA, Human Protein Atlas and Kaplan–Meier plotter databases. (OS: overall survival; NA: not applicable; RFS: recurrence-free survival). a OS curves of BRCA in TCGA. Low NEFM mRNA expression correlated with poor OS in TCGA_BRCA cohort (median OS: 149 vs. NA months, p 0.0017). b OS curves of BRCA in Human Protein Atlas database. NEFM protein = expression correlated with favorable OS (p 0.0014). c RFS curves of BRCA in Kaplan–Meier plotter databases (median RFS: 37.8 vs. 69.2 months, = p 1.5e-10). d OS curves of BRCA in Kaplan–Meier plotter databases (median OS: 88.67 vs. 143 months p 0.025). e–r. OS curves of pan_cancer = = in Kaplan–Meier plotter databases. e Cervical squamous cell carcinoma; f kidney renal clear cell carcinoma; g lung adenocarcinoma; h lung squamous cell carcinoma; i pancreatic ductal adenocarcinoma; j pheochromocytoma and paraganglioma; k rectum adenocarcinoma; l stomach adenocarcinoma; m thyroid carcinoma; n uterine corpus endometrial carcinoma; o bladder carcinoma; p head–neck squamous cell carcinoma; q ovarian cancer; r Sarcoma. High NEFM expression correlated with favorable OS of pancreatic ductal adenocarcinoma, pheochromocytoma and paraganglioma, whereas with poor OS of kidney renal clear cell carcinoma, lung adenocarcinoma, stomach adenocarcinoma, bladder carcinoma, head–neck squamous cell carcinoma, ovarian cancer and sarcoma in Kaplan–Meier plotter databases

Illumina HiSeq RNA-Seq dataset including 1110 tumors Lower NEFM expression (n = 545; MST: 149 months) was associated with worse OS, compared to higher versus 113 normal breast tissues (Fig. 3a). In addition, levels of 3 DNA methyltransferases were signif- expression (n = 544; MST: NA, and p = 0.0017) (Fig. 2a) in females in (TCGA) BRCA cohort, while NEFM pro- cantly diferent between NEFM high-expression group and low-expression group. Higher levels of DNMT1 tein expression correlated with favorable OS (p = 0.0014) in Human Protein Atlas database (Fig. 2b). In univari- (12.45 vs. 12.65), DNMT3A (11.41 vs. 11.57), DNMT3B able Cox proportional hazards regression models clini- (8.97 vs. 9.33) were observed in NEFM low-expression group (Fig. 3e–g, p < 0.001). Integrated analysis con- cal stage (p < 0.001), NEFM expression (p = 0.002), frmed the inverse relationship of NEFM DNA methyla- menopause (p = 0.04), age (p < 0.001), ER (p = 0.007), tion in Illumina Infnium HumanMethylation450 array PR (p = 0.013), HER2 (p = 0.021) were prognostic factors for OS. Furthermore, clinical stage (p < 0.001), age with NEFM transcriptional expression in TCGA breast tumors (Fig. 3h–j). In TCGA BRCA HumanMethyla- (p = 0.001), and NEFM expression (p = 0.031) remained as independent prognostic factors of OS in multivariable tion450K cohort, higher level of NEFM DNA methyla- Cox proportional hazards regression model (Table 2). To tion of NEFM was associated with poor OS (HR = 1.6 further explore prognostic potential of NEFM in tumors, p = 0.035) (Table 3). Six loci of NEFM DNA methylation Kaplan–Meier plotter derived from Afymetrix micro- were signifcantly associated with OS in BRCA based on arrays was applied. In accordance with TCGA BRCA, univariable Cox proportional hazards regression survival higher NEFM expression correlated with better progno- analysis (cg02761376, HR = 1.56, p = 0.045; cg07502389, sis of BRCA (OS: MST: 143 vs. 88.67 months for high vs. HR = 1.76, p = 0.012; cg09234518, HR = 1.96, p = 0.003; cg18267374, HR 1.75, p 0.013; cg19677607, HR 1.6, low NEFM expression, n = 1402, HR = 0.75, 95% CI 0.59– = = = p 0.038; cg26330518, HR 1.56, p 0.044). cg26330518 0.97, p = 0.025; recurrence-free survival (RFS): MST: 69.2 = = = is located in promoter N_Shore, the other fve loci are vs. 37.8 months, n = 3951, HR = 0.7, 95% CI 0.62–0.78, − located in promoter CpG island region. p = 1.5e 10) (Fig. 2c, d). Inverse correlation of NEFM DNA methylation with NEFM The genes and pathways connected with NEFM transcriptional expression transcriptional expression/DNA methylation Genome-wide DNA methylation array and gene expres- Diferentially expressed genes associated with NEFM sion profles of breast tissues from TCGA were explored expression or NEFM DNA methylation were profled to investigate the relationship of DNA methylation with through comparison between NEFM/NEFMmet high and transcriptional expression of NEFM. Methylation levels low groups in TCGA BRCA cohort. Totally, 164 up-reg- of NEFM were tested in Illumina Infnium HumanMeth- ulated and 546 down-regulated genes were signifcantly ylation450 array and Illumina Infnium HumanMeth- associated with NEFM expression, while 103 up-regu- ylation27 array based on 24 and 2 Infnium probes, lated and 641 down-regulated genes were signifcantly respectively, in 1103 tumors versus 123 normal breast associated with NEFM DNA methylation (with absolute tissues (788 tumors vs. 96 normal with HumanMethyla- value of log2foldchange > 1, and adjust p value < 0.05; tion450 array; 315 tumors vs. 27 normal with Human- Fig. 4a, d; Additional fles 1, 2: Tables S1–S2). Te top 50 Methylation27 array). Comparing with normal tissues, diferentially expressed genes were presented as expres- higher levels of NEFM DNA methylation of NEFM were sion heatmaps (Fig. 4b, e). Critical signal transduction observed in tumors (Fig. 3b–d), while NEFM transcrip- pathways involved in NEFM expression included neuro- tional expression was lower in tumor based on BRCA active ligand-receptor interaction, protein digestion and Li et al. Clin Epigenet (2021) 13:112 Page 5 of 18 Li et al. Clin Epigenet (2021) 13:112 Page 6 of 18

Table 2 Univariable and multivariable Cox proportional hazards regression models of NEFM expression with clinicopathological features in TCGA BRCA cohort Characteristics Univariate Multivariate HR CI95 p HR CI95 p

Subgroup 1.03 0.89–1.2 0.648 Stage 1.8 1.38–2.35 < 0.001 1.83 1.34–2.51 < 0.001 NEFM 0.53 0.36–0.8 0.002 0.58 0.36–0.95 0.031 Age 1.03 1.02–1.05 < 0.001 1.03 1.01–1.05 0.001 ER 0.56 0.37–0.85 0.007 0.62 0.3–1.28 0.198 PR 0.6 0.4–0.9 0.013 0.67 0.33–1.37 0.271 HER2 1.75 1.09–2.8 0.021 1.61 0.96–2.69 0.071

HR: hazard ratio; CI 95: 95% confdence interval; subgroup: luminal A, luminal B, positive HER2, basal, normal; stage: I, II, III, IV; NEFM: low or high expression by median. Age, ER, PR, HER2 was divided into two groups according to the median, respectively. Signifcant p values < 0.05 are in bold absorption, chemical carcinogenesis, cAMP signaling infltration in breast cancer. In TIMER gene modules, pathway, IL-17 signaling pathway, and cytokine–cytokine NEFM transcriptional expression positively correlated receptor interaction by KEGG enrichment analysis with infltration levels of CD8+ T cell, macrophage, neu- (Fig. 4c). Cytokine–cytokine receptor interaction, viral trophil, and dendritic cell, and negatively correlated with protein interaction with cytokine and cytokine receptor, infltration level of B cell and tumor purity, whereas not primary immunodefciency, hematopoietic cell lineage, with infltration level of CD4 + T cell. After adjusted chemokine signaling pathway, neuroactive ligand–recep- with tumor purity, NEFM expression weekly negatively tor interaction, T cell receptor signaling pathway, natural correlated with infltration level of B cell and positively killer cell-mediated cytotoxicity, IL-17 signaling path- correlated with macrophage and CD8+ T cell. NEFM way, and NF-kappa B signaling pathway were the top 10 DNA methylation was moderately to strongly (R > 3) pathways closely associated with NEFM DNA methyla- negatively associated with infltration levels of B cells, tion based on KEGG enrichment analysis. Notably, some CD8 + T cells, CD4 + T cells, macrophages, neutrophils, pathways involved in immune response such as T17 and dendritic cells using correlation analysis and TISIDB cell diferentiation, graft-versus-host disease, intestinal database (n = 785) (Fig. 5b, d). Interestingly, NEFM tran- immune network for IgA production, T1 and T2 cell scriptional expression weakly negatively correlated to diferentiation, as well as PD-L1 expression and PD-1 infltration levels of M2 macrophage, while NEFM DNA checkpoint pathway in cancer, were signifcantly associ- methylation weakly negatively correlated to infltra- ated with NEFM methylation (Fig. 4f). tion levels of M1 macrophage and positively correlated to infltration levels of M2 macrophage with correlation Correlation of NEFM transcription/DNA methylation analysis (Fig. 5e). Collectively, NEFM expression posi- with immune infltration in breast cancer tively correlated with macrophage infltration in TIMER Relationship of NEFM transcriptional expression/DNA and TISIDB; after adjusted with tumor purity, NEFM methylation with immune infltration in breast can- expression also weekly negatively correlated with infltra- cer was assessed using correlation analysis and TISIDB tion level of B cell and positively correlated with CD8+ databases. NEFM transcriptional expression was weakly T cell in TIMER gene modules. However, NEFM DNA (R < 2) to moderately (2 < R < 3) positively associated methylation was signifcantly negatively associated with with infltration levels of macrophages and neutrophils immune infltration in breast cancer. NEFM expression/ using correlation analysis and TISIDB database (Fig. 5a, DNA methylation might play a specifc role in immune c). NEFM transcriptional expression was weakly posi- infltration in BRCA. tively associated with infltration levels of CD8 + T cells, Correlation of NEFM transcriptional expression/DNA CD4 + T cells by correlation analysis, whereas weakly to moderately negatively associated with infltration lev- methylation with immune markers els of activated CD8 + T cells, activated CD4 + T cells in Relationship of NEFM transcriptional expression/ TISIDB database (Fig. 5a, c). Since the diferent results DNA methylation with immune markers was evalu- from correlation analysis and TISIDB databases, TIMER ated using TISIDB and TCGA databases. NEFM tran- gene modules were applied to evaluate the relation- scriptional expression weakly to moderately correlated ship of NEFM transcriptional expression with immune with ADORA2A, CSF1R, IDO1, KDR, LAG3, TGFBR1, Li et al. Clin Epigenet (2021) 13:112 Page 7 of 18

Fig. 3 The correlation between NEFM transcriptional expression and DNA methylation of NEFM in TCGA BRCA cohort. a NEFM transcriptional expression was downregulated in BRCA compared with normal breast tissue. b–d NEFM DNA methylation was enhanced in BRCA compared with normal breast tissue (b NEFMmet mean beta value of NEFM DNA methylation using 24 probe of HumanMethylation450K platform and 2 probe of = HumanMethylation27K; c NEFMmet mean beta value of NEFM DNA methylation of HumanMethylation450K platform; d NEFMmet mean beta = = value of NEFM DNA methylation of HumanMethylation27K platform). e–g Signifcant higher DNA methyltransferase was connected with NEFM low expression. e DNMT1 median expression 12.65 vs. 12.45 in NEFM low vs. high-expression group, p < 0.001; f DNMT3A median expression 11.57 vs. 11.41 in NEFM low vs. high-expression group, p < 0.001; g DNMT1 median expression 9.33 vs. 8.97 in NEFM low- vs. high-expression group, p < 0.001). h–j. DNA methylation of NEFM inversely correlated with NEFM expression. h NEFMmet log2 mean beta value of NEFM DNA methylation of 24 = probe of HumanMethylation450K platform, HR − 0.21, p 3.9e−09; i NEFMmet log2 beta value of cg17078116 probe, HR − 0.29, p < 2.2e−16; j = = = = NEFMmet log2 beta value of cg26980244 probe, HR − 0.30, p < 2.2e−16 = = Li et al. Clin Epigenet (2021) 13:112 Page 8 of 18

Table 3 Univariable Cox proportional hazards regression survival analyses of diferent NEFM DNA methylation loci in TCGA BRCA HumanMethylation450K platform VarNames UCSC_RefGene_Group Relation_to_UCSC_CpG_ HR CI95 p Island

NEFMmet 1.6 1.03–2.48 0.036 cg01583969 5’UTR;Body S_Shore 1.47 0.95–2.28 0.086 cg02002551 5’UTR;Body S_Shore 1.14 0.74–1.76 0.545 cg02106941 TSS1500;1stExon Island 0.97 0.63–1.49 0.888 cg02761376 TSS1500;1stExon Island 1.56 1.01–2.42 0.045 cg03012544 Body;Body S_Shore 1.14 0.74–1.75 0.555 cg03169018 TSS200;Body Island 1.2 0.78–1.84 0.412 cg04118306 TSS200;1stExon Island 0.97 0.63–1.49 0.885 cg07502389 TSS200;TSS1500 Island 1.76 1.14–2.74 0.012 cg07552803 TSS1500;1stExon Island 1.31 0.85–2.02 0.216 cg09234518 TSS200;TSS1500 Island 1.96 1.25–3.07 0.003 cg12026749 5’UTR;Body S_Shore 1.24 0.8–1.9 0.335 cg13387869 3’UTR;3’UTR S_Shelf 0.84 0.55–1.29 0.424 cg16459364 TSS200;TSS1500 Island 1.14 0.74–1.76 0.557 cg17078116 TSS200;1stExon Island 1.11 0.72–1.7 0.647 cg18267374 TSS1500;5’UTR;1stExon Island 1.75 1.13–2.72 0.013 cg18898125 TSS1500 N_Shore 1.22 0.79–1.87 0.368 cg19677607 TSS200;1stExon Island 1.6 1.03–2.48 0.038 cg20585869 TSS200;1stExon Island 0.78 0.5–1.2 0.261 cg22562942 TSS200;1stExon Island 1.35 0.87–2.1 0.181 cg23290344 TSS1500;1stExon Island 1.34 0.87–2.06 0.188 cg24705551 Body;Body S_Shelf 1.12 0.73–1.72 0.596 cg26330518 TSS1500 N_Shore 1.56 1.01–2.41 0.044 cg26980244 5’UTR;1stExon;Body Island 1.42 0.92–2.19 0.116 cg27475652 Body;Body S_Shelf 1.04 0.67–1.6 0.867

HR: hazard ratio; CI 95: 95% confdence interval. Signifcant p Values < 0.05 are in bold. NEFMmet: mean beta value of all DNA methylation loci of NEFM. Seven DNA methylation loci of NEFM including NEFMmet were connected with survival

VTCN1, C10orf54, CD276, CD40, CD70, ENTPD1, Discussion NT5E, PVR, TMEM173, TNFRSF13B, TNFRSF17, In this study, NEFM transcriptional expression was TNFSF4 (1 < R < 3) and strongly correlated with CXCL12 downregulated and negatively correlated with DNA and TGFB1 (R > 3). Except positive relation with PVRL2 methylation in breast cancer. Enhanced DNA meth- but no relation with CD276, RAET1E, TNFRSF14, ylation on six loci within NEFM located on promoter TNFRSF18, or TNFSF13, NEFM DNA methylation sig- CpG island or shore was associated with poor survival. nifcantly negatively correlated with almost all immu- Besides, NEFM transcriptional expression correlated nomodulators collected from Charoentong’s study with better prognosis and correlated with increased (Table 4). NEFM transcriptional expression was weakly macrophage; after normalized with tumor purity, NEFM associated with major histocompatibility complex expression correlated with increased CD8+ T cell, (MHC)-related molecules B2M, HLA-DPB1. Except whereas decreased B cell infltration in BRCA. NEFM TAPBP, NEFM DNA methylation was signifcantly nega- DNA methylation correlated with decreased infltra- tively associated with all MHC-related molecules listed tion levels of B cells, CD8 + T cells, CD4 + T cells, mac- in TISIDB database (Table 5). NEFM transcriptional rophages, neutrophils, and dendritic cells and diverse expression signifcantly positively correlated with CCL14, immune markers. Terefore, our study provides new evi- CCL21, CXCL12, CXCL14, CCL28, CCR10 and CX3CR1 dence to support a role of NEFM transcriptional expres- and negatively correlated with CCL7, CCL8, CCL18, sion/NEFM DNA methylation in BRCA. while NEFM DNA methylation was signifcantly nega- NEFM polypeptide is one of the four subunits compris- tively associated with most chemokines and receptors ing neuroflaments, the most abundant intermediate fla- listed in TISIDB (Fig. 6). ments in nervous system. In addition, NEFM, NEFL and Li et al. Clin Epigenet (2021) 13:112 Page 9 of 18

Fig. 4 Genes and pathways connected with NEFM transcriptional expression/DNA methylation in TCGA BRCA cohort. a Volcano plot of diferentially expressed gene profles between NEFM high-expression group and NEFM low-expression group (absolute log2 (fold change) > 1, adjp < 0.05). b Expression heatmap of the top 50 NEFM-associated genes. The top curve described NEFM’s expression distribution of 1096 BRCA samples. c Bar plot of the NEFM expression_related signaling pathways derived from KEGG Eenrichment analysis. d Volcano plot of diferentially expressed gene profles between NEFM high DNA methylation group and NEFM low DNA methylation group (absolute log2 (fold change) > 1, adjp < 0.05). e Expression heatmap of the top 50 NEFM DNA methylation-associated genes. The top curve described NEFM methylation distribution of 691 BRCA samples. f Bar plot of NEFM methylation-related signaling pathways derived from KEGG enrich analysis

NEFH act as onco-suppressors for afecting cell prolif- methylation-mediated inactivation of NEFH, NEFL and eration and correlate with worse prognosis [8, 31, 32]. It NEFM also occur in other types of cancer originated was reported that methylation-mediated inactivation of from pancreas, gastric and colon [8]. Consistently, we NEFH, NEFL or NEFM was common in primary breast demonstrated that NEFM transcriptional expression tumors compared to normal breast tissues and corre- was downregulated in most cancers including breast, lated with clinical features of disease progression. DNA colorectal, gastric, kidney, head and neck, compared Li et al. Clin Epigenet (2021) 13:112 Page 10 of 18

(See fgure on next page.) Fig. 5 Correlation of NEFM transcriptional expression/DNA methylation with immune infltration levels in BRCA. a NEFM transcriptional expression weakly positively correlated to infltration levels of CD8 T cells, CD4 T cells and macrophages in BRCA in TIMER database by correlation + + analysis. NEFM expression showed very weak positive association with infltration levels of neutrophils and dendritic cells in TCGA BRCA in TIMER database (n 703). b DNA methylation of NEFM had signifcant negative association with infltration levels of B cells, CD8 T cells, CD4 T cells, = + + macrophages, neutrophils, and dendritic cells in TIMER database by correlation analysis (n 703). c NEFM expression had weak positive association = with infltration levels of activated B cells, macrophages and moderate positive association with infltration levels of neutrophils, whereas negative association with infltration levels of activated CD8 T cells, activated CD4 T cells and activated dendritic cells in BRCA in TISIDB database + + (n 1100). d DNA methylation of NEFM signifcantly negatively correlated with infltration levels of activated B cells, activated CD8 T cells, = + activated CD4 T cells, macrophages, neutrophils, and activated dendritic cells in TISIDB database (n 785). e NEFM expression weakly negatively + = correlated with infltration levels of M2 macrophage; DNA methylation of NEFM negatively correlated with infltration levels of M1 macrophage whereas positively correlated with infltration levels of M2 macrophage in TIMER database (n 703). f NEFM transcriptional expression positively = correlated with infltration levels of CD8 T cells, macrophages, neutrophils, dendritic cells and negatively correlated with infltration level of B cells + in TIMER gene modules (n 1100). e After adjusted by tumor purity, NEFM expression weekly negatively correlated with infltration level of B cell = and positively correlated with infltration levels of macrophages and CD8 T cells in TIMER gene modules (n 1100) + =

with normal tissues in Oncomine and TIMER databases. module and TISIDB; however, relationship of NEFM NEFM DNA methylation negatively correlated with sur- expression with infltration levels of other TILs varied, vival and NEFM transcriptional expression in BRCA. In possibly due to diferent numbers of available TCGA our study, DNA methyltransferases (DNMT1, DNMT3A samples used for batch correction and diferences in and DNMT3B) were highly expressed in NEFM low- calculation methods. TIMER gene module (version 2.0) expression group in TCGA BRCA samples, suggesting provides abundance of immune infltration estimated by that DNMT1, DNMT3A and DNMT3B might contribute multiple immune deconvolution methods and adjusted to NEFM silencing in BRCA. Emerging evidence indi- by tumor purity, which is a major confounding factor cates that promoter methylation is associated with gene in this analysis, and thus, the results are more accurate, silencing, development, progression and chemotherapy with more reliable biological signifcance. Relationship of sensitivity of BRCA [2, 33, 34]. Terefore, the identi- NEFM transcriptional expression with immunomodula- fcation of novel tumor-suppressive genes targeted by tors has implicated its involvement in regulating tumor promoter methylation can reveal tumor-suppressive immunology in BRCA. Firstly, macrophage markers IL6, pathways in breast carcinogenesis and explore alterna- CSF1R, CXCL12 were weak-to-strong positively associ- tive approaches for diagnostic and therapeutic evalua- ated with NEFM expression, which could reveal a poten- tion. We investigated the relationship between 24 loci tial role of NEFM transcriptional expression in regulating within NEFM gene and prognosis in BRCA and identifed polarization of tumor-associated macrophage (TAM). enhanced NEFM DNA methylation of six out of 16 loci In addition, NEFM transcriptional expression was posi- located on promoter CpG island related to poor survival. tively associated with levels of T-cell exhaustion mark- Elevated levels of anti-NEFM antibodies were detected ers, specifcally ADORA2A, VISTA and CCR4 [38, 39]. in various neurological diseases, including autoim- Moreover, NEFM was positively associated with ecto- mune diseases, non-immune-mediated conditions, and nucleotidases CD39 and CD73, novel checkpoint inhibi- even in individuals being considered normal or with tors that interfere with anti-tumor immune responses disorders unrelated to intrathecal space, such as mul- [40]. NEFM negatively correlated with PVR (CD155), an tiple sclerosis, schizophrenia, spondylogenic headache immune checkpoint on tumor cells and interacting with or neurastenia. Terefore, anti-NEFM antibodies may CD96, CD226, and TIGIT (T cell immune receptor with be regarded as natural circulating auto-antibodies [35, immunoglobulin and ITIM domains) on TILs to modu- 36]. Poly-specifc T cells targeting distinct self-antigens late immune function in tumor microenvironment [41]. have been identifed in healthy individuals as well as in In addition, NEFM signifcantly positively correlated with the context of autoimmunity. T cell recognizes NEFM TGFB1, TGFBR1. Depending on the presence of other protein, with implications for aggravation and perpetua- secreted factors and cell surface co-receptors, TGFB can tion of central nervous system autoimmunity [37]. How- either suppress adaptive immune responses (through ever, whether NEFM is involved in regulating antitumor induction and stabilization of Tregs and directly sup- immunity with clinical signifcance in breast cancer pressing T1 cell, T2 cell and CD8 + T cell) or enhance remains unknown. In this study, positive association of adaptive responses (through induction of T17 cell, T9 NEFM expression with infltration level of macrophage cell and CD4 + CTL-like efector cell) [42]. Te relation- was replicated by correlation analysis, TIMER2.0 gene ship of NEFM with TILs in BRCA may partially rely on Li et al. Clin Epigenet (2021) 13:112 Page 11 of 18 Li et al. Clin Epigenet (2021) 13:112 Page 12 of 18

Table 4 Correlation of NEFM transcriptional expression/DNA methylation with immunomodulators based on TISIDB and TCGA database Immunomodulators NEFM expression NEFM DNA methylation TISIDB rho, n 1100 p TISIDB rho, n 785 P (TCGA R, n = 808) (TCGA R, n = 808) = = ADORA2A 0.137 (0.15) 5.07e 6 (2.5e 05) − 0.095 (− 0.004) 0.00751 (0.26) − − BTLA 0.032 (0.041) 0.296 (0.240) − 0.477 (− 0.35) 2.2e 16 − CD160 − 0.019 (0.07) 0.539 (0.046) − 0.157 (− 0.15) 9.61e 06 (1.5e 05) − − CD244 0.06 (0.076) 0.0467 (0.032) − 0.525 (− 0.35) < 2.2e 16 − CD274(PD-L1) − 0.006 (0.038) 0.854 (0.28) − 0.304 (− 0.28) 4.23e 18 (2.3e 15) − − CD96 0.037 (0.042) 0.226 (0.23) − 0.479 (− 0.34) < 2.2e 16 − CSF1R 0.132 (0.11) 0.22e 05 (0.0027) − 0.44 (− 0.34) < 2.2e 16 − − CTLA4 − 0.028 (0.00073) 0.357 (0.98) − 0.432 (0.3) < 2.2e 16 − HAVCR2 0.008 (− 0.0079) 0.789 (0.82) − 0.298 (− 0.21) 2.16e 17 (1.8e 09) − − IDO1 − 0.112 (− 0.1) 0.000203 (0.0036) − 0.384 (− 0.26) < 2.2e 16 (1.2e 13) − − IL10 − 0.003 (− 0.025) 0.921 (0.48) − 0.388 (− 0.29) < 2.2e 16 − IL10RB − 0.08 (− 0.036) 0.00821 (0.31) − 0.262 (− 0.2) 1.08e 13 (1.3e 08) − − KDR(VEGFR) 0.127 (0.11) 2.51e 05 (0.0017) − 0.117 (− 0.15) 0.000984 (2.9e 05) − − LAG3 − 0.171 (− 0.14) 1.11e 08 (4.5e 05) − 0.314 (− 0.2) 2.38e 19 (6.9e 09) − − − − LGALS9 − 0.051 (− 0.059) 0.0912 (0.093) − 0.293 (− 0.17) 9.15e 17 (1.2e 06) − − PDCD1 0.007 (0.023) 0.816 (0.51) − 0.43 (− 0.29) < 2.2e 16 − PDCD1LG2 0.066 (0.076) 0.0281 (0.031) − 0.476 (− 0.34) < 2.2e 16 − PVRL2(NECTIN2) 0.027 (0.045) 0.378 (0.2) 0.32 (0.14) 2.9e 20 (9.9e 05) − − TGFB1 0.32 (0.3) < 2.2e 16 − 0.28 (− 0.2)3) 1.91e 15 (3.6e 11) − − − TGFBR1 0.214 (0.21) 9.26e 13 (2.7e 09) − 0.178 (− 0.13) 5.46e 07 (0.00014) − − − TGFBR2 (0.35) < 2.2e 16 (− 0.33) < 2.2e 16 − − TIGIT − 0.02 (− 0.0034) 0.509 (0.92) − 0.446 (− 0.31) < 2.2e 16 − VTCN1 0.207 (0.21) 4.34e 12 (2.4e 09) − 0.175 (− 0.07) 8.63e 07 (0.048) − − − C10orf54(VSIR, VISTA) 0.205 (0.18) 7.97e 12 (2.3e 07) − 0.552 (− 0.39) < 2.2e 16 − − − CD27(TNFRSF7) 0.097 ( (0.11) 0.00121 (0.0013) − 0.452 (− 0.31) < 2.2e 16 − CD276 0.168 (0.19) 2.12e 08 (9.7e 08) − 0.038 (− 0.065) 0.284 (0.068) − − CD28 0.042 (0.038) 0.16 (0.28) − 0.458 (− 0.33) < 2.2e 16 − CD40 0.137 (0.15) 5.1e 06 (2.2e 05) − 0.553 (− 0.38) < 2.2e 16 − − − CD40LG 0.088 (0.093) 0.00346 (0.0081) − 0.457 (0.32) < 2.2e 16 − CD48 0.062 (0.062) 0.0391 (0.079) − 0.458 (− 0.31) < 2.2e 16 − CD70 0.108 (0.11) 0.000345 (0.0017) − 0.312 (− 0.17) 3.39e 19 (1.4e 06) − − CD80 − 0.069 0.0228 − 0.259 (− 0.18) 1.95e 13 (2.8e 07) − − CD86 − 0.028 (− 0.056) 0.357 (0.11) − 0.375 < 2.2e 16 − CXCL12 0.467 (0.43) < 2.2e 16 − 0.32 (− 0.2) 3.31e 20 (1.5e 08) − − − CXCR4 0.053 (0.087) 0.0763(0.013) − 0.313 (− 0.17) 3.11e 19 (8.2e 07) − − ENTPD1(CD39) 0.267 (0.3) 2.85e 19 (< 2.2e 16) − 0.271 (− 0.23) 1.46e 14 (5.5e 11) − − − − ICOS − 0.064 (− 0.044) 0.0341 (0.21) − 0.42 (− 0.29) < 2.2e 16 − ICOSLG 0.045 (0.0042) 0.15 (0.91) − 0.193 (− 0.027) 5.69e 08 (0.44) − IL2RA − 0.033 (− 0.022) 0.276 (0.53) − 0.454 (− 0.33) < 2.2e 16 − IL6 0.091 (0.087) 0.00266 (0.014) − 0.425 (− 0.3) < 2.2e 16 − IL6R − 0.012 (− 0.015) 0.7 (0.67) − 0.341 (− 0.29) < 2.2e 16 − KLRC1 0.004 (0.017) 0.907 (0.64) − 0.386 (− 0.28) < 2.2e 16 (2.7e 16) − − KLRK1 0.041 (0.075) 0.171 (0.033) − 0.456 (− 0.28) < 2.2e 16 (4.3e 16) − − LTA − 0.033 (− 0.015) 0.273 (0.66) − 0.413 (− 0.28) < 2.2e 16 (1e 15) − − MICB 0.003 (− 0.0094) 0.93 (0.79) − 0.047 (− 0.16) 0.186(3e 06) − NT5E(CD73) 0.224 (0.2) 6.85e 14 (1e 08) − 0.306 (− 0.28) 2.26e 18 (5.6e 16) − − − − Li et al. Clin Epigenet (2021) 13:112 Page 13 of 18

Table 4 (continued) Immunomodulators NEFM expression NEFM DNA methylation TISIDB rho, n 1100 p TISIDB rho, n 785 P (TCGA R, n = 808) (TCGA R, n = 808) = = PVR − 0.204 (− 0.17) 1.05e 11 (8.9e 07) − 0.134 (− 0.11) 0.00017 (0.0019) − − RAET1E 0.094 (0.12) 0.00182 (0.00065) − 0.073 (− 0.016) 0.0411 (0.65) TMEM173(STING) 0.231 (0.2) 1.14e 14 (6.5e 09) − 0.263 (− 0.19) 9.54e 14 (1.1e 07) − − − − TNFRSF13B 0.202 (0.21) 1.39e 11 (3.8e 09) − 0.419 (− 0.31) < 2.2e 16 − − − TNFRSF13C 0.068 (0.027) 0.0237 (0.44) − 0.346 (− 0.26) < 2.2e 16 (4.6e 14) − − TNFRSF14 0.048 (0.055) 0.109 (0.12) − 0.074 (0.044) 0.0377 (0.21) TNFRSF17 0.113 (0.13) 0.000181(0.00037) − 0.396 (− 0.27) < 2.2e 16 (3.9e 15) − − TNFRSF18 − 0.029 (− 0.04) 0.335 (0.25) 0.064 (− 0.021) 0.0727 (0.00037) TNFRSF25 0.014 (0.078) 0.646 (0.028) − 0.257 (− 0.14) 3.01e 13 (4.3e 05) − − TNFRSF4 0.082 (0.082) 0.00664 (0.019) − 0.281 (− 0.18) 1.51e 15 (2.9e 07) − − TNFRSF8 0.086 (0.085) 0.00424 (0.015) − 0.529 (− 0.37) < 2.2e 16 − TNFRSF9 0.03 (0.036) 0.319 (0.3) − 0.443 (− 0.36) < 2.2e 16 − TNFSF13 − 0.015 (0.028) 0.617 (0.42) 0.047 (0.075) 0.189 (0.035) TNFSF13B − 0.015 (− 0.025) 0.623 (0.48) − 0.343 (− 0.25) < 2.2e 16 (1.3e 12) − − TNFSF14 0.082 (0.084) 0.00622 (0.017) − 0.45 (− 0.32) < 2.2e 16 − TNFSF15 0.095 (0.085) 0.00153 (0.015) − 0.201 (− 0.12) 1.36e 08 (0.00091) − TNFSF4 0.149 (0.17) 6.77e 07 (2.1e 06) − 0.081 (− 0.096) 0.0228 (0.0064) − − TNFSF9 0.024 (0.086) 0.00165 (0.015) − 0.256 (− 0.11) 3.82e 13 (0.0012) − ULBP1 (NKG2D) − 0.029 (0.0054) 0.333 (0.88) − 0.159 (− 0.015) 7.37e 06 (0.68) − Signifcant p value < 0.05 is in bold chemokines and chemokine receptors. More and more structure regulation, are critical for tumor microenvi- studies have shown that chemokines and chemokine ronment (TME) (including immune cells) interaction. receptors are closely related to the immunity of breast Emerging evidence supports that tumors commonly cancer. NEFM transcriptional expression was negatively hijack various methylation mechanisms to escape associated with CCL7, CCL8 and CCL18, which would immune surveillance. Recent studies have identifed a recruit monocytes to diferentiate into tumor-associated strong connection between epigenetics and cytokine macrophages (TAMs) at the tumor site, indicating that production in tumorigenesis [49, 50]. Methyltrans- NEFM may cause decreased M2 macrophage infltra- ferases regulate production of interferons, cytokines and tion [43]. Furthermore, NEFM transcriptional expres- chemokines [51]. KMT3A (SETD2), a methyltransferase, sion was positively associated with CCL14, CCL21, is required for interferon pathway by catalyzing the CXCL12, CXCL14, CCL28, CCR10 and CX3CR1. methylation of STAT1, a key transcription factor of inter- Notably, CX3CR1 promotes macrophage recruitment feron response [52]. DNMT suppresses MHC-I expres- during mammary tumor formation. Macrophages are sion on tumor cells [53]. Te upregulation of PD-L1 on attracted to tumor sites expressing chemotactic factors tumor cells likely results from selection pressure exerted such as CCL7, CCL8 and CXCL12 [43, 44]. Addition- by T cell immune response. Epigenetic mechanisms cer- ally, CXCL12 promotes neutrophil infltration to tumors. tainly contribute to upregulation of PDL1 [54]. Meth- Moreover, CXCL12 is a potent attractant of dendritic ylation regulators KMT6A (EZH2), MBD2, TET2 and cells (DCs); CCL21 recruits DCs and regulatory T cells demethylase KDM5B have been implicated in lympho- (Tregs) [45]. CCL14 participates in the infltration of the cyte development [55]. DNMT3A controls fate decision tumor by anti-cancer TILs. CXCL14 is responsible for of early efector CD8 + T cell. Loss of DNMT3A leads to immune cell recruitment and maturation and is critical inefective repression of genes that are supposed to be to upregulating major histocompatibility complex class I silenced in efector cells, thus generating fewer efector expression on tumor cells. CCL28 activates CCR10 and cells [56]. Tese studies have revealed special relationship causes B cell and T cell migration [46–48]. between TME-infltrating immune cells and DNA meth- Epigenetic mechanisms, including DNA methylation, ylation modifcation, beyond RNA degradation. NEFM histone posttranslational modifcations and chromatin DNA methylation signifcantly negatively correlated with Li et al. Clin Epigenet (2021) 13:112 Page 14 of 18

Table 5 Correlation of NEFM transcriptional expression/DNA include cytokine–cytokine receptor interaction, viral methylation with MHC molecules based on TISIDB database protein interaction with cytokine and cytokine recep- MHC NEFM expression NEFM DNA methylation tor, chemokine signaling pathway, natural killer cell- molecules mediated cytotoxicity, primary immunodefciency, T cell TISIDB rho, p TISIDB rho, p n 1100 n 785 receptor signaling pathway, IL-17 signaling pathway and = = PD-L1 expression and PD-1 checkpoint pathway in can- B2M − 0.11 0.00025 − 0.287 2.85e 16 − cer, which are signifcantly associated with NEFM DNA HLA-A − 0.093 0.00205 − 0.3 1.23e 17 − methylation by KEGG enrichment analysis. HLA-B − 0.089 0.00388 − 0.315 1.58e 19 − It has been reported that interrogation of site-specifc HLA-C − 0.077 0.0107 − 0.212 2.35e 09 − CpG sites may be another option for assessing immune HLA-DMA 0.058 0.0561 − 0.405 < 2.2e 16 − infltration in tumors and may possibly predict response HLA-DMB 0.03 0.313 − 0.414 < 2.2e 16 − to checkpoint inhibitors [58]. Since NEFM DNA meth- HLA-DOA 0.156 2.16e 07 − 0.458 < 2.2e 16 − − ylation signifcantly negatively correlated with TILs and HLA-DOB 0.034 0.266 − 0.503 < 2.2e 16 − many immune pathways (especially PD-L1 expression HLA-DPA1 0.084 0.00511 − 0.407 < 2.2e 16 − and PD-1 checkpoint pathway in cancer) in BRCA, the HLA-DPB1 0.153 3.46e 07 − 0.435 < 2.2e 16 − − six CpG sites within NEFM promoter associated with HLA-DQA1 0.061 0.0446 − 0.398 < 2.2e 16 − poor prognosis may serve as biomarkers for predict- HLA-DQA2 0.057 0.058 − 0.294 5.78e 14 − ing immune infltration in BRCA. Our results also sug- HLA-DQB1 0.081 0.00747 − 0.352 < 2.2e 16 − gest that demethylation of NEFM might be a strategy HLA-DRA 0.061 0.0447 − 0.434 < 2.2e 16 − to improve the efciency of immunotherapy. Tus, it is HLA-DRB1 0.089 0.00329 − 0.39 < 2.2e 16 − required to further explore the detailed mechanism and HLA-E 0.05 0.0991 − 0.484 < 2.2e 16 − function of transcriptional expression/DNA methylation HLA-F − 0.028 0.358 − 0.377 < 2.2e 16 − in regulating tumor microenvironment. HLA-G − 0.04 0.189 − 0.231 6.32e 11 − TAP1 − 0.204 8.63e 12 − 0.267 3.39e 14 Conclusion − − TAP2 − 0.176 4.89e 09 − 0.381 < 2.2e 16 − − NEFM transcriptional expression positively correlates TAPBP − 0.143 1.91e 06 − 0.057 0.109 − with favorable prognosis and increased levels of mac- Signifcant p value < 0.05 is in bold rophage infltration in BRCA. After adjusted by tumor purity, NEFM expression correlates with increased infltration of CD8+ T cell, whereas decreased infltration various immunomodulators and most chemokines and of B cell. NEFM DNA methylation correlates with poor receptors listed in TISIDB, which also contributed to prognosis and decreased immune infltration of B cells, decreased immune cells infltration in BRCA. CD8 + and CD4 + T cells, macrophages, neutrophils M1 macrophages are involved in normal T1 immune and dendritic cells in BRCA. Moreover, NEFM expres- responses, whereas M2 macrophages support survival sion/DNA methylation correlates with diverse immune and dissemination of cancer cells via secretion of various markers and pathways in BRCA. Terefore, our study factors, including cytokines, chemokines and enzymes, highlights potential clinical signifcance of NEFM tran- which recruit Tregs intratumorally to suppress antitu- scriptional expression/DNA methylation in breast cancer mor cytotoxicity [57]. We demonstrated that NEFM tran- and provides insight into a novel role of NEFM expres- scriptional expression was signifcantly associated with sion/DNA methylation in tumor immune infltration. IL-17 signaling pathway and cytokine–cytokine receptor interaction by KEGG enrichment analysis. In breast Methods cancer, we infer that NEFM transcriptional expression BRCA data and sources may afect survival partially through the decreased M2 TCGA BRCA DNA methylation profles (Infnium macrophage infltration and anti-tumor cytotoxicity of HumanMethylation450K and HumanMethylation27K) cytokine–cytokine receptor interaction and IL-17 signal- of 1226 breast tissues (884 HumanMethylation450K and ing pathway. However, molecular mechanisms should be 342 HumanMethylation27K samples) and gene expres- further investigated. One potential mechanism by which sion profles (IlluminaHiSeq_RNA-SeqV2) of 1223 breast NEFM methylation associated with poor survival may tissues as well as clinical were downloaded from TCGA be NEFM methylation inducing tumor immunosuppres- data repository (https://tcga-data.nci.nih.gov/tcga/). sion depending on decreased TILs. Other mechanisms Ten, data of immune infltrates in BRCA were down- underlying relationship of NEFM DNA methylation with loaded from TIMER database (http://timer.cistrome. immune infltration and poor prognosis in BRCA may org/). Li et al. Clin Epigenet (2021) 13:112 Page 15 of 18

Fig. 6 The correlation of NEFM transcriptional expression/DNA methylation with chemokines and receptors in TISIDB database. a Correlation of NEFM transcriptional expression with chemokines; b correlation of DNA methylation of NEFM with chemokines; c correlation of NEFM expression with chemokine receptors; d correlation of DNA methylation of NEFM with chemokine receptors

Expression levels of NEFM in various types of cancer used to evaluate impacts of NEFM expression on OS in Te expression profling of NEFM in various types of the presence of other known risk factors. NEFM associ- cancer was identifed in the Oncomine database (https:// ated with OS and RFS of BRCA patients was validated in www.oncomine.org/resource/login.html) [26]. Te Kaplan–Meier Plotter database (http://kmplot.com/analy threshold was determined according to the following val- sis/) [27] among 5,143 BRCA patients. Potential efects of ues: p-value of 0.001, fold change of 1.5, and gene rank- NEFM expression on OS were evaluated in Pan-cancer ing of all. We also analyzed NEFM expression in diferent RNA-seq in Kaplan–Meier plotter database. Te asso- types of cancer in TIMER database. ciation of NEFM protein expression with BRCA OS was analyzed by Te Human Protein Atlas database (http:// Prognosis assessment www.proteinatlas.org/). Te Kaplan–Meier plotter and univariable Cox proportional hazards regression models were used to estimate Immune infltration association between NEFM transcriptional expression or The data of immune infiltration in BRCA were down- DNA methylation and median survival time. Multivari- loaded from TIMER database. Relationship of NEFM able Cox proportional hazards regression models were transcriptional expression or DNA methylation with Li et al. Clin Epigenet (2021) 13:112 Page 16 of 18

the abundance of immune infiltration was evaluated, analyzed with DESeq2 R package. Volcano plots and including B cells, CD4 + T cells, CD8 + T cells, neu- heatmaps were presented. KEGG (Kyoto Encyclopedia trophils, macrophages, M1 macrophages, M2 mac- of Genes and Genomes) enrichment analyses were per- rophages and dendritic cells by R package. We also formed with R package to identify pathways related to analyzed the relationship of NEFM expression with NEFM transcriptional expression/NEFM DNA methyl- the abundance of immune infiltration using gene mod- ation. A p-value of < 0.05 was considered as statistically. ules in TIMER. Gene expression level normalized with tumor purity was displayed on the leftmost panel. Statistical analyses There were 28 TIL elements in TISIDB database. Overall survival (OS) was calculated from the date of Relationship of NEFM transcriptional expression/ diagnosis to death due to any causes or to last follow- DNA methylation with abundance of TILs (including up. Recurrence-free survival (RFS) was calculated activated CD8+ T cell, activated CD4+ T cell, activated from the date of diagnosis to local relapse/recurrence dendritic cell, activated B cell, macrophage, and neu- or regional relapse/recurrence or death (all causes) trophil) was examined in TISIDB database. whichever occurs frst. Te Kaplan–Meier method and In addition, the relationship of NEFM transcrip- log-rank test were used to estimate the relationship tional expression or DNA methylation with immune of NEFM transcriptional expression/DNA methyla- gene markers was explored via Spearman’s correlation with OS and RFS. Te Fisher exact and Wilcoxon tion. These immune gene markers included immu- rank-sum tests were used, respectively, for categori- nomodulators collected from Charoentong’s study cal and continuous variables, to assess the relationship [28], chemokines and receptors based on TISIDB data- of NEFM expression levels and clinical or molecular base [29]. The correlation scatter plots between NEFM characteristics. Multivariable Cox proportional haz- transcriptional expression/DNA methylation and ards regression models were used to evaluate potential immune infiltration levels of immune cells, in BRCA, impact of NEFM expression on OS in the presence of together with Spearman’s correlation and estimated other known risk factors. Student’s t-test and multiple statistical significance, were described. The log2RSEM hypothesis correction (false discovery rate, FDR) were value of NEFM expression and log2β value of NEFM used to identify diferences in genome-wide genes, DNA methylation were used for x-axis, whereas methylation profles between NEFM high and NEFM low related immune infiltration levels of immune cells for groups. Spearman correlation analysis was performed y-axis. Specific levels of gene markers were displayed to evaluate the relationship of NEFM methylation with with log2RSEM. transcriptional expression or other genes. A p value of TISIDB (http://cis.hku.hk/TISIDB/index.php) is a less than 0.05 was considered statistically signifcant. web portal for tumor and immune system interaction, All analyses were performed using R 3.6.1 software which integrates multiple heterogeneous data sets. packages. Te relative abundance of TILs as demonstrated by 28-gene immune-related signature from Charoentong’s Abbreviations study was estimated by using gene set variation analy- NEFM: Neuroflament medium; BRCA: Breast cancer; NEFH: Neuroflament sis (GSVA) based on gene expression profle in TISIDB heavy; NEFL: Neuroflament light; oAECs: Ovine amniotic epithelium; MDD: Major depressive disorder; LOH: Loss of heterozygosity; DRIP: Dopamine database (Additional fle 3) [29]. receptor-interacting protein; OS: Overall survival; DFS: Recurrence-free survival; TIMER [30] is a comprehensive resource for system- DNMT1: DNA methyltransferase 1; DNMT3A: DNA methyltransferase 3 alpha; atic analysis of immune infltrates across diverse cancer DNMT3B: DNA methyltransferase 3 beta; ADORA2A: Adenosine A2a receptor; CSF1R: Colony-stimulating factor 1 receptor; IDO1: Indoleamine 2,3-dioxyge- types (https://cistrome.shinyapps.io/timer/). TIMER nase 1; KDR: Kinase insert domain receptor; LAG3: Lymphocyte-activating 3; applies a deconvolution, a previously published sta- TGFB1: Transforming growth factor beta 1; TGFBR1: Transforming growth factor tistical method, to infer relative abundance of tumor- beta receptor 1; VTCN1: V-set domain containing T cell activation inhibitor 1; VSIR (C10ORF54): V-set immunoregulatory receptor; CD276: CD276 molecule; infltrating immune cells from gene expression profles. CD40: CD40 molecule; CD70: CD70 molecule; ENTPD1 (CD39): Ectonucleoside TIMER database includes 10,897 samples across 32 triphosphate diphosphohydrolase 1; NT5E (CD73): 5’-Nucleotidase ecto; PVR cancer types from Te Cancer Genome Atlas (TCGA) (CD155): PVR cell adhesion molecule; TMEM173: Transmembrane protein 173; TNFRSF13B: TNF receptor superfamily member 13B; TNFRSF17: TNF receptor to estimate relative abundance of immune infltration. superfamily member 17; TNFSF4: TNF superfamily member 4; CXCL12: C-X-C motif chemokine ligand 12; TGFB1: Transforming growth factor beta 1; HTN1: Histatin 1; MAGEA10: MAGE family member A10; CRISPLD2: Cysteine-rich Enrichment analysis secretory protein LCCL domain containing 2; HTR7: 5-Hydroxytryptamine receptor 7; FST: Follistatin; ASPN: Aspirin; FLRT2: Fibronectin leucine-rich Diferentially expressed genes associated with NEFM transmembrane protein 2; CPXM1: Carboxypeptidase X, M14 family member expression and levels of NEFM DNA methylation were 1; SCN4B: Sodium voltage-gated channel beta subunit 4; PTN: Pleiotrophin; PAK5: P21 (RAC1)-activated kinase 5; CCR4: C–C motif chemokine receptor 4; Li et al. Clin Epigenet (2021) 13:112 Page 17 of 18

CD96: CD96 molecule; CD226: CD226 molecule; TIGIT: T cell immunoreceptor 2. Jeschke J, Bizet M, Desmedt C, Calonne E, Dedeurwaerder S, with Ig and ITIM domains; TAM: Tumor-associated macrophages; Tregs: Regula- Garaud S, et al. DNA methylation-based immune response signa- tory T cells; TILs: Tumor-infltrating lymphocytes. ture improves patient diagnosis in multiple cancers. J Clin Invest. 2017;127(8):3090–102. 3. Ignatiadis M, Singhal S, Desmedt C, Haibe-Kains B, Criscitiello C, Supplementary Information Andre F, et al. Gene modules and response to neoadjuvant chemo- The online version contains supplementary material available at https://doi. therapy in breast cancer subtypes: a pooled analysis. J Clin Oncol. org/10.1186/s13148-021-01096-4. 2012;30(16):1996–2004. 4. Adams S, Gatti-Mays M, Kalinsky K, Korde L, Sharon E, Amiri-Kordestani L, et al. Current landscape of immunotherapy in breast cancer: a review. Additional fle 1. The diferentially expressed genes between NEFM high- JAMA Oncol. 2019;5(8):1205–14. expression group and NEFM low-expression group. 5. Schmid P, Adams S, Rugo H, Schneeweiss A, Barrios C, Iwata H, et al. Additional fle 2. The diferentially expressed genes between NEFM high- Atezolizumab and nab-paclitaxel in advanced triple-negative breast methylation group and NEFM low-methylation group. cancer. N Engl J Med. 2018;379(22):2108–21. 6. Denkert C, Loibl S, Noske A, Roller M, Müller B, Komor M, et al. Tumor- Additional fle 3. 28 tumor-infltrating lymphocytes (TILs) and related associated lymphocytes as an independent predictor of response to neo- gene signatures in TISIDE database. adjuvant chemotherapy in breast cancer. J Clin Oncol. 2010;28(1):105–13. 7. Wein L, Savas P, Luen S, Virassamy B, Salgado R, Fio Loi SJ. Clinical validity and utility of tumor-infltrating lymphocytes in routine clinical practice Acknowledgements for breast cancer patients: current and future directions. Front Oncol. The authors gratefully acknowledge contributions of Professor Yunyan Gu’s 2017;7:1856–64. research team from the Department of Bioinformatics of Harbin Medical 8. Calmon M, Jeschke J, Zhang W, Dhir M, Siebenkäs C, Herrera A, et al. University in bioinformatics information, of Dr. Xiaoqun Dong from the Warren Epigenetic silencing of neuroflament genes promotes an aggressive Alpert Medical School, Brown University in editing English language, and of Dr. phenotype in breast cancer. Epigenetics. 2015;10(7):622–32. Ruifen Zhang from Institute of Microbiology, Chinese Academy of Sciences for 9. Kudo L, Parfenova L, Vi N, Lau K, Pomakian J, Valdmanis P, et al. Integrative an interpretation of this paper’s results. gene-tissue microarray-based approach for identifcation of human disease biomarkers: application to amyotrophic lateral sclerosis. Human Authors’ contributions Mol Genetics. 2010;19(16):3233–53. Dandan Li and Wehhao Zhao performed the experiments, interpreted data, 10. Skvortsova V, Shadrina M, Slominsky P, Levitsky G, Kondratieva E, Zherebt- analyzed data, and drafted the manuscript. Xinyu Zhang interpreted data and sova A, et al. Analysis of heavy neuroflament subunit gene polymor- edited the manuscript. Hanning Lv interpreted data. Lichun Sun and Chun- phism in Russian patients with sporadic motor neuron disease (MND). hong Li designed and supervised the study, interpreted data, and edited the EJHG. 2004;12(3):241–4. manuscript. All authors read and approved the fnal manuscript. 11. Mersiyanova I, Perepelov A, Polyakov A, Sitnikov V, Dadali E, Oparin R, et al. A new variant of Charcot-Marie-Tooth disease type 2 is probably the Funding result of a mutation in the neuroflament-light gene. Am J Human Genet- This study was supported by The National Cancer Center climbing Fund No. ics. 2000;67(1):37–46. NCC201808B22, Heilongjiang Provincial Health Commission scientifc research 12. Bergson C, Levenson R, Goldman-Rakic P, Lidow M. Dopamine receptor- project No. 2019-066 and National Natural Science Foundation Grant of China interacting proteins: the Ca(2 ) connection in dopamine signaling. No. 81573001. Trends Pharmacol Sci. 2003;24(9):486–92.+ 13. Strous R, Greenbaum L, Kanyas K, Merbl Y, Horowitz A, Karni O, et al. Asso- Availability of data and materials ciation of the dopamine receptor interacting protein gene, NEF3, with All data generated or analyzed during this study are included in this published early response to antipsychotic medication. Int J Neuropsychopharmacol. article and its supplementary information fles. 2007;10(3):321–33. 14. Alholle A, Brini A, Gharanei S, Vaiyapuri S, Arrigoni E, Dallol A, et al. Func- Declarations tional epigenetic approach identifes frequently methylated genes in Ewing sarcoma. Epigenetics. 2013;8(11):1198–204. Ethics approval and consent to participate 15. Huang Z, Zhuo Y, Shen Z, Wang Y, Wang L, Li H, et al. The role of NEFL in Not applicable. cell growth and invasion in head and neck squamous cell carcinoma cell lines. J Oral Pathol Med. 2014;43(3):191–8. Consent for publication 16. Revill K, Wang T, Lachenmayer A, Kojima K, Harrington A, Li J, et al. Not applicable. Genome-wide methylation analysis and epigenetic unmasking identify tumor suppressor genes in hepatocellular carcinoma. Gastroenterology. Competing interests 2013;145(6):1424–35. Not applicable. 17. Emi M, Fujiwara Y, Nakajima T, Tsuchiya E, Tsuda H, Hirohashi S, et al. Frequent loss of heterozygosity for loci on chromosome 8p in hepa- Author details tocellular carcinoma, colorectal cancer, and lung cancer. Cancer Res. 1 Department of Radiotherapy Oncology, Harbin Medical University Cancer 1992;52(19):5368–72. Hospital, Harbin, China. 2 Department of Breast Medical Oncology, Harbin 18. Vogelstein, B,E Fearon, S Kern, S Hamilton, A Preisinger, Y Nakamura, , et al. Medical University Cancer Hospital, No.150 Haping Road, Nangang District, Allelotype of colorectal carcinomas. Science. 1989;244(4901):207–11. Harbin 150081, China. 19. Yaremko M, Kutza C, Lyzak J, Mick R, Recant W, Westbrook C. Loss of heterozygosity from the short arm of chromosome 8 is associated Received: 15 July 2020 Accepted: 2 May 2021 with invasive behavior in breast cancer. Genes Chromosomes Cancer. 1996;16(3):189–95. 20. Hagihara A, Miyamoto K, Furuta J, Hiraoka N, Wakazono K, Seki S, et al. Identifcation of 27 5’ CpG islands aberrantly methylated and 13 genes silenced in human pancreatic cancers. Oncogene. 2004;23(53):8705–10. References 21. Kim M, Chang X, LeBron C, Nagpal J, Lee J, Huang Y, et al. Neuroflament 1. Esteva F, Hubbard-Lucey V, Tang J, Pusztai L. Immunotherapy and tar- heavy polypeptide regulates the Akt-beta-catenin pathway in human geted therapy combinations in metastatic breast cancer. Lancet Oncol. esophageal squamous cell carcinoma. PLoS ONE. 2010;5(2):e9003. 2019;20(3):175–86. 22. Dubrowinskaja N, Gebauer K, Peters I, Hennenlotter J, Abbas M, Scherer R, et al. Neuroflament heavy polypeptide CpG island methylation Li et al. Clin Epigenet (2021) 13:112 Page 18 of 18

associates with prognosis of renal cell carcinoma and prediction of signifcance and exhausted tumor-infltrating lymphocytes: a cohort antivascular endothelial growth factor therapy response. Cancer Med. study. J Immunol Res. 2020;2020:3948928. 2014;3(2):300–9. 42. Guo Q, Betts C, Pennock N, Mitchell E, Schedin P. Mammary gland involu- 23. Tabarés-Seisdedos R, Rubenstein J. Chromosome 8p as a potential hub tion provides a unique model to study the TGF-β cancer paradox. J Clin for developmental neuropsychiatric disorders: implications for schizo- Med. 2017;6(1):10. phrenia, autism and cancer. Mol Psychiatry. 2009;14(6):563–89. 43. Do H, Lee C, Cho J. Chemokines and their receptors: multifaceted roles 24. Barboni B, Russo V, Curini V, Martelli A, Berardinelli P, Mauro A, et al. in cancer progression and potential value as cancer prognostic markers. Gestational stage afects amniotic epithelial cells phenotype, methylation Cancers. 2020;12(2):287. status, immunomodulatory and stemness properties. Stem Cell Rev Rep. 44. Schmall A, Al-Tamari H, Herold S, Kampschulte M, Weigert A, Wietelmann 2014;10(5):725–41. A, et al. Macrophage and cancer cell cross-talk via CCR2 and CX3CR1 is a 25. Lucca L, Axisa P, Aloulou M, Perals C, Ramadan A, Rufas P, et al. Myelin fundamental mechanism driving lung cancer. Am J Respir Crit Care Med. oligodendrocyte glycoprotein induces incomplete tolerance of CD4( ) T 2015;191(4):437–47. cells specifc for both a myelin and a neuronal self-antigen in mice. Eur+ J 45. Mollica Poeta V, Massara M, Capucetti A, Bonecchi R. Chemokines and Immunol. 2016;46(9):2247–59. chemokine receptors: new targets for cancer immunotherapy. Front 26. Rhodes D, Kalyana-Sundaram S, Mahavisno V, Varambally R, Yu J, Briggs Immunol. 2019;10:379. B, et al. Oncomine 3.0: genes, pathways, and networks in a collection of 46. Westrich J, Vermeer D, Colbert P, Spanos W, Pyeon D. The multifarious 18,000 cancer gene expression profles. Neoplasia. 2007;9(2):166–80. roles of the chemokine CXCL14 in cancer progression and immune 27. Györfy B, Lanczky A, Eklund A, Denkert C, Budczies J, Li Q, et al. An online responses. Mol Carcinog. 2020;59(7):794–806. survival analysis tool to rapidly assess the efect of 22,277 genes on breast 47. Korbecki J, Kojder K, Simińska D, Bohatyrewicz R, Gutowska I, Chlubek D, cancer prognosis using microarray data of 1,809 patients. Breast Cancer et al. CC chemokines in a tumor: a review of pro-cancer and anti-cancer Res Treat. 2010;123(3):725–31. properties of the ligands of receptors CCR1, CCR2, CCR3, and CCR4. Int J 28. Charoentong P, Finotello F, Angelova M, Mayer C, Efremova M, Rieder Mol Sci. 2020;21(21):8412. D, et al. Pan-cancer immunogenomic analyses reveal genotype-immu- 48. Korbecki J, Grochans S, Gutowska I, Barczak K, Baranowska-Bosiacka I. CC nophenotype relationships and predictors of response to checkpoint chemokines in a tumor: a review of pro-cancer and anti-cancer proper- blockade. Cell Rep. 2017;18(1):248–62. ties of receptors CCR5, CCR6, CCR7, CCR8, CCR9, and CCR10 ligands. Int J 29. Ru B, Wong C, Tong Y, Zhong J, Zhong S, Wu W, et al. TISIDB: an integrated Mol Sci. 2020;21(20):7619. repository portal for tumor-immune system interactions. Bioinformatics. 49. McDonald K, Kawaguchi T, Qi Q, Peng X, Asaoka M, Young J, et al. Tumor 2019;35(20):4200–2. heterogeneity correlates with less immune response and worse survival 30. Li T, Fan J, Wang B, Traugh N, Chen Q, Liu J, et al. TIMER: a web server for in breast cancer patients. Ann Surg Oncol. 2019;26(7):2191–9. comprehensive analysis of tumor-infltrating immune cells. Cancer Res. 50. Roulois D, Loo Yau H, Singhania R, Wang Y, Danesh A, Shen S, et al. DNA- 2017;77(21):108–10. demethylating agents target colorectal cancer cells by inducing viral 31. Chen K, Liu J, Liu S, Xia M, Zhang X, Han D, et al. Methyltransferase SETD2- mimicry by endogenous transcripts. Cell. 2015;162(5):961–73. mediated methylation of STAT1 is critical for interferon antiviral activity. 51. Chiappinelli K, Strissel P, Desrichard A, Li H, Henke C, Akman B, et al. Inhib- Cell. 2017;170(3):492–506. iting DNA methylation causes an interferon response in cancer via dsRNA 32. Peng G, Yuan X, Yuan J, Liu Q, Dai M, Shen C, et al. miR-25 promotes including endogenous retroviruses. Cell. 2015;162(5):974–86. glioblastoma cell proliferation and invasion by directly targeting NEFL. 52. Peng D, Kryczek I, Nagarsheth N, Zhao L, Wei S, Wang W, et al. Epigenetic Mol Cell Biochem. 2015;409(1–2):103–11. silencing of TH1-type chemokines shapes tumour immunity and immu- 33. Ricketts C, Morris M, Gentle D, Shuib S, Brown M, Clarke N, et al. Methyla- notherapy. Nature. 2015;527(7577):249–53. tion profling and evaluation of demethylating therapy in renal cell 53. Li H, Chiappinelli K, Guzzetta A, Easwaran H, Yen R, Vatapalli R, et al. carcinoma. Clin Epigenetics. 2013;5(1):16. Immune regulation by low doses of the DNA methyltransferase inhibi- 34. Pasculli B, Barbano R, Parrella P. Epigenetics of breast cancer: biology and tor 5-azacitidine in common human epithelial cancers. Oncotarget. clinical implication in the era of precision medicine. Semin Cancer Biol. 2014;5(3):587–98. 2018;51:22–35. 54. Zhu H, Bengsch F, Svoronos N, Rutkowski M, Bitler B, Allegrezza M, et al. 35. Bartos A, Fialová L, Soukupová J, Kukal J, Malbohan I, Pit’ha J. Elevated BET bromodomain inhibition promotes anti-tumor immunity by sup- intrathecal antibodies against the medium neuroflament subunit in pressing PD-L1 expression. Cell Rep. 2016;16(11):2829–37. multiple sclerosis. J Neurol. 2007;254(1):20–5. 55. Cao J, Yan Q. Cancer epigenetics, tumor immunity, and immunotherapy. 36. Jones A, Mowry B, McLean D, Mantzioris B, Pender M, Greer J. Elevated Trends Cancer. 2020;6(7):580–92. levels of autoantibodies targeting the M1 muscarinic acetylcholine 56. Ladle B, Li K, Phillips M, Pucsek A, Haile A, Powell J, et al. De novo DNA receptor and neuroflament medium in sera from subgroups of patients methylation by DNA methyltransferase 3a controls early efector CD8 with schizophrenia. J Neuroimmunol. 2014;269:68–75. T-cell fate decisions following activation. Proc Natl Acad Scie USA. + 37. Lucca L, Desbois S, Ramadan A, Ben-Nun A, Eisenstein M, Carrié N, et al. 2016;113(38):10631–6. Bispecifcity for myelin and neuronal self-antigens is a common feature of 57. Solinas G, Germano G, Mantovani A, Allavena P. Tumor-associated mac- CD4 T cells in C57BL/6 mice. J Immunol. 2014;193(7):3267–77. rophages (TAM) as major players of the cancer-related infammation. J 38. Zhang S, Wang Y, Gu Y, Zhu J, Ci C, Guo Z, et al. Specifc breast cancer Leukocyte Biol. 2009;86(5):1065–73. prognosis-subtype distinctions based on DNA methylation patterns. Mol 58. Bacolod M, Barany F, Fisher P. Can CpG methylation serve as sur- Oncol. 2018;12(7):1047–60. rogate markers for immune infltration in cancer? Adv Cancer Res. 39. Zarour H. Reversing T-cell Dysfunction and Exhaustion in Cancer. Clin 2019;143:351–84. Cancer Res. 2016;22(8):1856–64. 40. Allard B, Longhi M, Robson S, Stagg J. The ectonucleotidases CD39 and CD73: novel checkpoint inhibitor targets. Immunol Rev. Publisher’s Note 2017;276(1):121–44. Springer Nature remains neutral with regard to jurisdictional claims in pub- 41. Li Y, Zhou Q, Song Q, Wang R, Lyu S, Guan X, et al. Overexpression of an lished maps and institutional afliations. immune checkpoint (CD155) in breast cancer associated with prognostic